the wideband screenshot (HDSDR Waterfall)
shows the typical signature of a switchmode power supply. Most choppers run at
lower frequency (eg 45 kHz), with the harmonic corrupting the LF
band. The carrier is FM-modulated by the 100 Hz ripple of he primary
voltage and usually swept across several kHz (which ironically is said to
help passing EMC tests). The banded structure is due to the beating
between the upward and downward chirps, a maximum occurs when the lag
between the two is such that they are in phase. Near the edges of the
spectrum, the bands are often wider because the two passes are closer in time
near the point of return.
The other spectrogram (137700...2130) looked a bit weird at first. But it
can be explained by the NB periodically triggering on the hash, twice per 10 ms.
This apparently creates 200 Hz aliases of the DCF39 signal, which are
visible as discrete lines on 138830 - N x 200 Hz. Of course
these lines also carry an image of the DCF modulation
telegrams.
Try to observe whether on and off times are correlated to
neighbours' christmas light effects (these often radiate badly due to long
and outdoor wiring). Or walk around and try to locate it with a portable LF
receiver in AM mode, preferably one with a small telescopic E-field
antenna.
A couple of years back I have made an attempt on a software hash
canceller, which estimates the repeating pattern of the chirps and tries to
subtract it. If you would like to experiment with this, it's at
Best 73,
Markus (DF6NM)